Blog
Notes for engineers evaluating agent skills, MCP servers, and install-time policy.
- November 5, 2026 · KillerSkillsInstructions your agent reads and you never seeUnicode tag characters, invisible HTML, poisoned MCP tool descriptions and injected tool results: where instructions hide from the reader, and what KillerSkills checks now at install time, in the sandbox and inside real sessions.
- November 5, 2026 · KillerSkillsAgent Lure Stealer: 51 fake Claude skill and MCP repos, scannedA campaign ships infostealer ZIPs inside repos posing as Claude skills, Codex tools and MCP servers. The SKILL.md is clean; the attack is in the README. What text scanners miss, what we changed, and a live scan of every repo.
- October 27, 2026 · KillerSkillsWhy the agent skill supply chain breaks (and how KillerSkills fixes it)Agent skills and MCP servers sit inside the agent trust boundary. Why allowlists and static scans fall short, and how install-time honeypot verdicts fix it.